Experts: Which Ones Should You Trust?kutrov.web.elte.hu/tudhat/11_goldman.pdf · 2014-05-15 ·...

Philosophy and Phenomenological Research Vol. LXIII, No. 1, July 2001

Experts: Which Ones Should You Trust?

ALVIN I. GOLDMAN

University of Arizona

1. Expertise and Testimony

Mainstream epistemology is a highly theoretical and abstract enterprise.

Traditional epistemologists rarely present their deliberations as critical to the

practical problems of life, unless one supposes-as Hume, for example, did

not-that skeptical worries should trouble us in our everyday affairs. But

some issues in epistemology are both theoretically interesting and practically

quite pressing. That holds of the problem to be discussed here: how lay-

persons should evaluate the testimony of experts and decide which of two or

more rival experts is most credible. It is of practical importance because in a

complex, highly specialized world people are constantly confronted with situ-

ations in which, as comparative novices (or even ignoramuses), they must

turn to putative experts for intellectual guidance or assistance. It is of theoret-

ical interest because the appropriate epistemic considerations are far from

transparent; and it is not clear how far the problems lead to insurmountable

skeptical quandaries. This paper does not argue for flat-out skepticism in this

domain; nor, on the other hand, does it purport to resolve all pressures in the

direction of skepticism. It is an exploratory paper, which tries to identify

problems and examine some possible solutions, not to establish those solu-

tions definitively.

The present topic departs from traditional epistemology and philosophy of

science in another respect as well. These fields typically consider the

prospects for knowledge acquisition in "ideal" situations. For example, epis-

temic agents are often examined who have unlimited logical competence and

no significant limits on their investigational resources. In the present prob-

lem, by contrast, we focus on agents with stipulated epistemic constraints

and ask what they might attain while subject to those constraints.

Although the problem of assessing experts is non-traditional in some

respects, it is by no means a new problem. It was squarely formulated and

addressed by Plato in some of his early dialogues, especially the Charmides.

In this dialogue Socrates asks whether a man is able to examine another man

who claims to know something to see whether he does or not; Socrates

wonders whether a man can distinguish someone who pretends to be a doctor

from someone who really and truly is one (Charmides 170d-e). Plato's term

for posing the problem is techne', often translated as "knowledge" but perhaps

better translated as "expertise" (see Gentzler 1995, LaBarge 1997).'

In the recent literature the novicelexpert problem is formulated in stark

terms by John Hardwig (1985, 1991). When a layperson relies on an expert,

that reliance, says Hardwig, is necessarily blind.2 Hardwig is intent on deny-

ing full-fledged skepticism; he holds that the receiver of testimony can acquire

"knowledge" from a source. But by characterizing the receiver's knowledge as

"blind", Hardwig seems to give us a skepticism of sorts. The term "blind"

seems to imply that a layperson (or a scientist in a different field) cannot be

rationally justi$ed in trusting an expert. So his approach would leave us with

testimonial skepticism concerning rational justification, if not knowledge.

There are other approaches to the epistemology of testimony that lurk in

Hardwig's neighborhood. The authors I have in mind do not explicitly urge

any form of skepticism about testimonial belief; like Hardwig, they wish to

expel the specter of skepticism from the domain of testimony. Nonetheless,

their solution to the problem of testimonial justification appeals to a mini-

mum of reasons that a hearer might have in trusting the assertions of a

source. Let me explain who and what I mean.

The view in question is represented by Tyler Burge (1993) and Richard

Foley (1994), who hold that the bare assertion of a claim by a speaker gives a

hearer prima facie reason to accept it, quite independently of anything the

hearer might know or justifiably believe about the speaker's abilities,

circumstances, or opportunities to have acquired the claimed piece of knowl-

edge. Nor does it depend on empirically acquired evidence by the hearer, for

example, evidence that speakers generally make claims only when they are in

a position to know whereof they speak. Burge, for example, endorses the

following Acceptance Principle: "A person is entitled to accept as true some-

thing that is presented as true and that is intelligible to him, unless there are

stronger reasons not to do so" (1993: 467). He insists that this principle is

not an empirical one; the "justificational force of the entitlement described by

this justification is not constituted or enhanced by sense experiences or

perceptual beliefs" (1993: 469). Similarly, although Foley does not stress the

a priori status of such principles, he agrees that it is reasonable of people to

grant fundamental authority to the opinions of others, where this means that

it is "reasonable for us to be influenced by others even when we have no

special information indicating that they are reliable" (1994: 55). Fundamental

Thanks to Scott LaBarge for calling Plato's treatment of this subject to my attention.

In his 1991 paper, Hardwig at first says that trust must be "at least partially blind" (p.

693). He then proceeds to talk about knowledge resting on trust and therefore being blind

(pp. 693,699) without using the qualifier "partially".

86 ALVIN I. GOLDMAN

authority is contrasted with derivative authority, where the latter is generated

from the hearer's reasons for thinking that the source's "information, abili-

ties, or circumstances put [him] in an especially good position" to make an

accurate claim (1994: 55). So, on Foley's view, a hearer need not have such

reasons about a source to get prima facie grounds for trusting that source.

Moreover, a person does not need to acquire empirical reasons for thinking

that people generally make claims about a subject only when they are in a

position to know about that subject. Foley grants people a fundamental

(though prima facie) epistemic right to trust others even in the absence of any

such empirical e ~ i d e n c e . ~ It is in this sense that Burge's and Foley's views

seem to license "blind" trust.

I think that Burge, Foley, and others are driven to these sorts of views in

part by the apparent hopelessness of reductionist or inductivist alternatives.

Neither adults nor children, it appears, have enough evidence from their

personal perceptions and memories to make cogent inductive inferences to the

reliability of testimony (cf. Coady 1992). So Burge, Foley, Coady and others

propose their "fundamental" principles of testimonial trustworthiness to stem

the potential tide of testimonial skepticism. I am not altogether convinced

that this move is necessary. A case might be made that children are in a

position to get good inductive evidence that people usually make claims

about things they are in a position to know about.

A young child's earliest evidence of factual reports is from face-to-face

speech. The child usually sees what the speaker is talking about and sees that

the speaker also sees what she is talking about, e.g., the furry cat, the toy

under the piano, and so forth. Indeed, according to one account of cognitive

development (Baron-Cohen 1995), there is a special module or mechanism,

the "eye-direction detector", that attends to other people's eyes, detects their

direction of gaze, and interprets them as "seeing" whatever is in the line of

sight.4 Since seeing commonly gives rise to knowing, the young child can

determine a certain range of phenomena within the ken of speakers. Since the

earliest utterances the child encounters are presumably about these speaker-

known objects or events, the child might easily conclude that speakers

usually make assertions about things within their ken. Of course, the child

later encounters many utterances where it is unclear to the child whether the

However, there is some question whether Foley can consistently call the epistemic right

he posits a "fundamental" one, since he also says that it rests on (A) my justified sey-

tmst, and (B) the similarity of others to me-presumably the evidence I have of their

similarity to me (see pp. 63-64). Another question for Foley is how the fundamentality

thesis fits with his view that in cases of conflict I have more reason (prima facie) to trust

myself than to trust someone else (see p. 66). If my justified tmst in others is really

fundamental, why does it take a back-seat to self-tmst?

Moreover, according to Baron-Cohen, there is a separate module called the "shared

attention mechanism", which seeks to determine when another person is attending to the

same object as the self is attending to.

EXPERTS: WHICH ONES SHOULD YOU TRUST? 87

-- - - -

matters reported are, or ever were, within the speaker's ken. Nonetheless, a

child's early experience is of speakers who talk about what they apparently

know about, and this may well be a decisive body of empirical evidence

available to the child.

I don't want to press this suggestion very hard.5 I shall not myself be

offering a full-scale theory about the justification of testimonial belief. In

particular, I do not mean to be advancing a sustained defense of the reduction-

ist or inductivist position. Of greater concern to me is the recognition that a

hearer's evidence about a source's reliability or unreliability can often bolster

or defeat the hearer's justifiedness in accepting testimony from that source.

This can be illustrated with two examples.

As you pass someone on the street, he assertively utters a sophisticated

mathematical proposition, which you understand but have never previously

assessed for plausibility. Are you justified in accepting it from this stranger?

Surely it depends partly on whether the speaker turns out to be a mathematics

professor of your acquaintance or, say, a nine-year-old child. You have prior

evidence for thinking that the former is in a position to know such a proposi-

tion, whereas the latter is not. Whether or not there is an a priori principle of

default entitlement of the sort endorsed by Burge and Foley, your empirical

evidence about the identity of the speaker is clearly relevant. I do not claim

that Burge and Foley (etc.) cannot handle these cases. They might say that

your recognition that the speaker is a math professor bolsters your overall

entitlement to accept the proposition (though not your prima facie entitle-

ment); recognizing that it is a child defeats your prima facie entitlement to

accept the proposition. My point is, however, that your evidence about the

properties of the speaker is crucial evidence for your overall entitlement to

accept the speaker's assertion. A similar point holds in the following exam-

ple. As you relax behind the wheel of your parked car, with your eyes closed,

you hear someone nearby describing the make and color of the passing cars.

Plausibly, you have prima facie justification in accepting those descriptions

as true, whether this prima facie entitlement has an a priori or inductivist

basis. But if you then open your eyes and discover that the speaker is himself

blindfolded and not even looking in the direction of the passing traffic, this

prima facie justification is certainly defeated. So what you empirically deter-

mine about a speaker can make a massive difference to your overall justified-

ness in accepting his utterances.

The same obviously holds about two putative experts, who make conflict-

ing claims about a given subject-matter. Which claim you should accept (if

either) can certainly be massively affected by your empirical discoveries about

For one thing, it may be argued that babies' interpretations of what people say is, in the

first instance, constrained by the assumption that the contents concern matters within the

speakers' perceptual ken. This is not an empirical finding, it might be argued, but an a

prion posit that is used to fix speakers' meanings.

88 ALVIN I. GOLDMAN

their respective abilities and opportunities to know the truth of the matter

(and to speak sincerely about it). Indeed, in this kind of case, default princi-

ples of the sort advanced by Burge and Foley are of no help whatever.

Although a hearer may be prima facie entitled to believe each of the speakers,

he cannot be entitled all things considered to believe both of them; for the

propositions they assert, we are supposing, are incompatible (and tsanspar-

ently incompatible to the hearer). So the hearer's all-things-considered justi-

fiedness vis-a-vis their claims will depend on what he empirically leans

about each speaker, or about the opinions of other speakers. In the rest of this

paper I shall investigate the kinds of empirical evidence that a novice hearer

might have or be able to obtain for believing one putative expert rather than

her rival. I do not believe that we need to settle the "foundational" issues in

the general theory of testimony before addressing this issue. This is the work-

ing assumption, at any rate, on which I shall p r~ceed .~

2. The Novice/Expert Problem vs. the ExpertLExpert Problem

There are, of course, degrees of both expertise and novicehood. Some novices

might not be so much less knowledgeable than some experts. Moreover, a

novice might in principle be able to t u n himself into an expert, by improv-

ing his epistemic position vis-a-vis the target subject-matter, e.g., by acquir-

ing more formal training in the field. This is not a scenario to be considered

in this paper, however. I assume that some sorts of limiting factors-whether

they be time, cost, ability, or what have you-will keep our novices from

becoming experts, at least prior to the time by which they need to make their

judgment. So the question is: Can novices, while remaining novices, make

justified judgments about the relative credibility of rival experts? When and

how is this possible?

There is a significant difference between the novice/expert problem and

another type of problem, the experttexpert problem. The latter problem is one

Some theorists of testimony, Burge included, maintain that a hearer's justificational status

vis-8-vis a claim received from a source depends partly on the justificational status of the

source's own belief in that claim. This is a transpersonal, preservationist, or transmissional

conception of justifiedness, under which a recipient is not justified in believing p unless

the speaker has a justification and entitlement that he transmits to the hearer. For

purposes of this paper, however, I shall not consider this transmissional conception of

justification. First, Burge himself recognizes that there is such a thing as the recipient's

"proprietary" justification for believing an interlocutor's claim, justification localized "in"

the recipient, which isn't affected by the source's justification (1993: 485-486). I think it is

appropriate to concentrate on this "proprietary" justification (of the recipient) for present

purposes. When a hearer is trying to "choose" between the conflicting claims of rival

speakers, he cannot appeal to any inaccessible justification lodged in the heads of the

speakers. He can only appeal to his own justificational resources. (Of course, these might

include things said by the two speakers by way of defense of their contentions, things

which also are relevant to their own justifications.) For other types of (plausible) objec-

tions to Burge's preservationism about testimony, see Bezuidenhout (1998).


in which experts seek to appraise the authority or credibility of other experts.

Philip Kitcher (1993) addresses this problem in analyzing how scientists

ascribe authority to their peers. A crucial segment of such authority ascrip-

tion involves what Kitcher calls "calibration" (1993: 314-22). In direct

calibration a scientist uses his own opinions about the subject-matter in

question to evaluate a target scientist's degree of authority. In indirect calibra-

tion, he uses the opinions of still other scientists, whose opinions he has

previously evaluated by direct calibration, to evaluate the target's authority.

So here too he starts from his own opinions about the subject-matter in

question.

By contrast, in what I am calling the novicelexpert problem (more specifi-

cally, the novicel2-expert problem), the novice is not in a position to evalu-

ate the target experts by using his own opinion; at least he does not think he

is in such a position. The novice either has no opinions in the target domain,

or does not have enough confidence in his opinions in this domain to use

them in adjudicating or evaluating the disagreement between the rival experts.

He thinks of the domain as properly requiring a certain expertise, and he does

not view himself as possessing this expertise. Thus, he cannot use opinions

of his own in the domain of expertise--call it the E-domain-to choose

between conflicting experts' judgments or reports.

We can clarify the nature of the novicetexpert problem by comparing it to

the analogous listenerleyewitness problem. (Indeed, if we use the term

"expert" loosely, the latter problem may just be a species of the novicelexpert

problem.) Two putative eyewitnesses claim to have witnessed a certain crime.

A listener-for example, a juror-did not himself witness the crime, and has

no prior beliefs about who committed it or how it was committed. In other

words, he has no personal knowledge of the event. He wants to learn what

transpired by listening to the testimonies of the eyewitnesses. The question is

how he should adjudicate between their testimonies if and when they conflict.

In this case, the E-domain is the domain of propositions concerning the

actions and circumstances involved in the crime. This E-domain is what the

listener (the "novice") has no prior opinions about, or no opinions to which

he feels he can legitimately appeal. (He regards his opinions, if any, as mere

speculation, hunch, or what have you.)

It may be possible, at least in principle, for a listener to make a reason-

able assessment of which eyewitness is more credible, even without having

or appealing to prior opinions of his own concerning the E-domain. For

example, he might obtain evidence from others as to whether each putative

witness was really present at the crime scene, or, alternatively, known to be

elsewhere at the time of the crime. Second, the listener could learn of tests of

each witness's visual acuity, which would bear on the accuracy or reliability

of their reports. So in this kind of case, the credibility of a putative "expert's"

report can be checked by such methods as independent verification of whether

90 ALVIN I. GOLDMAN

he had the opportunity and ability to see what he claims to have seen. Are

analogous methods available to someone who seeks to assess the credibility

of a "cognitive" expert as opposed to an eyewitness expert?

Before addressing this question, we should say more about the nature of

expertise and the sorts of experts we are concerned with here. Some kinds of

experts are unusually accomplished at certain skills, including violinists,

billiards players, textile designers, and so forth. These are not the kinds of

experts with which epistemology is most naturally concerned. For epistemo-

logical purposes we shall mainly focus on cognitive or intellectual experts:

people who have (or claim to have) a superior quantity or level of knowledge

in some domain and an ability to generate new knowledge in answer to ques-

tions within the domain. Admittedly, there are elements of skill or know-how

in intellectual matters too, so the boundary between skill expertise and cogni-

tive expertise is not a sharp one. Nonetheless, I shall try to work on only one

side of this rough divide, the intellectual side.

How shall we define expertise in the cognitive sense? What distinguishes

an expert from a layperson, in a given cognitive domain? I'll begin by speci-

fying an objective sense of expertise, what it is to be an expert, not what it is

to have a reputation for expertise. Once the objective sense is specified, the

reputational sense readily follows: a reputational expert is someone widely

believed to be an expert (in the objective sense), whether or not he really is

one.

Turning to objective expertise, then, I first propose that cognitive exper-

tise be defined in "veritistic" (truth-linked) terms. As a first pass, experts in a

given domain (the E-domain) have more beliefs (or high degrees of belief) in

true propositions and/or fewer beliefs in false propositions within that

domain than most people do (or better: than the vast majority of people do).

According to this proposal, expertise is largely a comparative matter. How-

ever, I do not think it is wholly comparative. If the vast majority of people

are full of false beliefs in a domain and Jones exceeds them slightly by not

succumbing to a few falsehoods that are widely shared, that still does not

make him an "expert" (from a God's-eye point of view). To qualify as a

cognitive expert, a person must possess a substantial body of truths in the

target domain. Being an expert is not simply a matter of veritistic superiority

to most of the community. Some non-comparative threshold of veritistic

attainment must be reached, though there is great vagueness in setting this

threshold.

Expertise is not all a matter of possessing accurate information. It

includes a capacity or disposition to deploy or exploit this fund of informa-

tion to form beliefs in true answers to new questions that may be posed in

the domain. This arises from some set of skills or techniques that constitute

part of what it is to be an expert. An expert has the (cognitive) know-how,

when presented with a new question in the domain, to go to the right sectors


of his information-bank and perform appropriate operations on this informa-

tion; or to deploy some external apparatus or data-banks to disclose relevant

material. So expertise features a propensity element as well as an element of

actual attainment.

A third possible feature of expertise may require a little modification in

what we said earlier. To discuss this feature, let us distinguish the primary

and secondary questions in a domain. Primary questions are the principal

questions of interest to the researchers or students of the subject-matter.

Secondary questions concern the existing evidence or arguments that bear on

the primary questions, and the assessments of the evidence made by promi-

nent researchers. In general, an expert in a field is someone who has

(comparatively) extensive knowledge (in the weak sense of knowledge, i.e.,

true belief) of the state of the evidence, and knowledge of the opinions and

reactions to that evidence by prominent workers in the field. In the central

sense of "expert" (a strong sense), an expert is someone with an unusually

extensive body of knowledge on both primary and secondary questions in the

domain. However, there may also be a weak sense of "expert", in which it

includes someone who merely has extensive knowledge on the secondary

questions in the domain. Consider two people with strongly divergent views

on the primary questions in the domain, so that one of them is largely right

and the other is largely wrong. By the original, strong criterion, the one who

is largely wrong would not qualify as an expert. People might disagree with

this as the final word on the matter. They might hold that anyone with a

thorough knowledge of the existing evidence and the differing views held by

the workers in the field deserves to be called an expert. I concede this by

acknowledging the weak sense of "expert".

Applying what has been said above, we can say that an expert (in the

strong sense) in domain D is someone who possesses an extensive fund of

knowledge (true belief) and a set of skills or methods for apt and successful

deployment of this knowledge to new questions in the domain. Anyone pur-

porting to be a (cognitive) expert in a given domain will claim to have such a

fund and set of methods, and will claim to have true answers to the ques-

t ion(~) under dispute because he has applied his fund and his methods to the

question(s). The task for the layperson who is consulting putative experts,

and who hopes thereby to learn a true answer to the target question, is to

decide who has superior expertise, or who has better deployed his expertise to

the question at hand. The novicet2-experts problem is whether a layperson

can justifiably choose one putative expert as more credible or trustworthy

than the other with respect to the question at hand, and what might be the

epistemic basis for such a ~ h o i c e ? ~

In posing the question of justifiedness, I mean to stay as neutral as possible between

different approaches to the concept of justifiedness, e.g., between internalist versus

3. Argument-Based Evidence

To address these issues, I shall begin by listing five possible sources of

evidence that a novice might have, in a novicel2-experts situation, for trust-

ing one putative expert more than another. I'll then explore the prospects for

utilizing such sources, depending on their availability and the novice's exact

circumstance. The five sources I shall discuss are:

(A) Arguments presented by the contending experts to support their

own views and critique their rivals' views.

(B) Agreement from additional putative experts on one side or other

of the subject in question.

(C) Appraisals by "meta-experts" of the experts' expertise (including

appraisals reflected in formal credentials earned by the experts).

(D) Evidence of the experts' interests and biases vis-a-vis the ques-

tion at issue.

(E) Evidence of the experts' past "track-records"

In the remainder of the paper, I shall examine these five possible sources,

beginning, in this section, with source (A).8

There are two types of communications that a novice, N, might receive

from his two experts, E, and E,.' First, each expert might baldly state her

view (conclusion), without supporting it with any evidence or argument

whatever. More commonly, an expert may give detailed support to her view

in some public or professional context, but this detailed defense might only

appear in a restricted venue (e.g., a professional conference or journal) that

does not reach N's attention. So N might not encounter the two experts'

defenses, or might encounter only very truncated versions of them. For

example, N might hear about the experts' views and their support from a

second-hand account in the popular press that does not go into many details.

At the opposite end of the communicational spectrum, the two experts might

engage in a full-scale debate that N witnesses (or reads a detailed reconstruc-

tion of). Each expert might there present fairly developed arguments in

externalist approaches to justifiedness. Notice, moreover, that I am not merely asking

whether and how the novice can justifiably decide to accept one (candidate) expert's

view outright, but whether and how he can justifiably decide to give greater credence to

one than to the other.

I do not mean to be committed to the exhaustiveness of this list. The list just includes some

salient categories. ' In what follows I shall for brevity speak about two experts, but I shall normally mean two

putative experts, because from the novice's epistemic perspective it is problematic

whether each, or either, of the self-proclaimed experts really is one.


support of her view and against that of her opponent. Clearly, only when N

somehow encounters the experts' evidence or arguments can he have evidence

of type (A). So let us consider this scenario.

We may initially suppose that if N can gain (greater) justification for

believing one expert's view as compared with the other by means of their

arguments, the novice must at least understand the evidence cited in the

experts' arguments. For some domains of expertise and some novices,

however, even a mere grasp of the evidence may be out of reach. These are

cases where N is an "ignoramus" vis-a-vis the E-domain. This is not the

universal plight of novices. Sometimes they can understand the evidence (in

some measure) but aren't in a position, from personal knowledge, to give it

any credence. Assessing an expert's evidence may be especially difficult when

it is disputed by an opposing expert.

Not every statement that appears in an expert's argument need be epistem-

ically inaccessible to the novice. Let us distinguish here between esoteric and

exoteric statements within an expert's discourse. Esoteric statements belong

to the relevant sphere of expertise, and their truth-values are inaccessible to

N-in terms of his personal knowledge, at any rate. Exoteric statements are

outside the domain of expertise; their truth-values may be accessible to N-

either at the time of their assertion or later."' I presume that esoteric state-

ments comprise a hefty portion of the premises and "lemmas" in an expert's

argument. That's what makes it difficult for a novice to become justified in

believing any expert's view on the basis of arguments per se. Not only are

novices commonly unable to assess the truth-values of the esoteric proposi-

tions, but they also are ill-placed to assess the support relations between the

cited evidence and the proffered conclusion. Of course, the proponent expert

will claim that the support relation is strong between her evidence and the

conclusion she defends; but her opponent will commonly dispute this. The

novice will be ill-placed to assess which expert is in the right.

At this point I wish to distinguish direct and indirect argumentative justi-

fication. In direct argumentative justification, a hearer becomes justified in

believing an argument's conclusion by becoming justified in believing the

argument's premises and their (strong) support relation to the conclusion. If a

speaker's endorsement of an argument helps bring it about that the hearer has

such justificational status vis-a-vis its premises and support relation, then the

hearer may acquire "direct" justification for the conclusion via that speaker's

--- -~

"' It might be helpful to distinguish semanticallp esoteric statements and eprsremlcallp

esoteric statements. (Thanks to Carol Caraway for this suggestion.) Semantically esotenc

statements are ones that a novice cannot assess because he does not even understand

them; typically, they utilize a technical vocabulary he has not mastered. Epistemically

esoteric statements are statements the novice understands but still cannot assess for truth-

value.

94 ALVIN I. GOLDMAN

argument.ll As we have said, however, it is difficult for an expert's argument

to produce direct justification in the hearer in the novicel2-expert situation.

Precisely because many of these matters are esoteric, N will have a hard time

adjudicating between El's and E,'s claims, and will therefore have a hard time

becoming justified vis-a-vis either of their conclusions. He will even have a

hard time becoming justified in trusting one conclusion more than the other.

The idea of indirect argumentative justification arises from the idea that

one speaker in a debate may demonstrate dialectical superiority over the other,

and this dialectical superiority might be a plausible indicator12 for N of greater

expertise, even if it doesn't render N directly justified in believing the supe-

rior speaker's conclusion. By dialectical superiority, I do not mean merely

greater debating skill. Here is an example of what I do mean.

Whenever expert E, offers evidence for her conclusion, expert El presents

an ostensible rebuttal or defeater of that evidence. On the other hand, when El

offers evidence for her conclusion, E, never manages to offer a rebuttal or

defeater to El's evidence. Now N is not in a position to assess the truth-value

of El ' s defeaters against E,, nor to evaluate the truth-value or strength of

support that El's (undefeated) evidence gives to El 's conclusion. For these

reasons, El ' s evidence (or arguments) are not directly justificatory for N.

Nonetheless, in "formal" dialectical terms, El seems to be doing better in the

dispute. Furthermore, I suggest, this dialectical superiority may reasonably be

taken as an indicator of El 's having superior expertise on the question at

issue. It is a (non-conclusive) indicator that El has a superior fund of informa-

tion in the domain, or a superior method for manipulating her information, or

both.

Additional signs of superior expertise may come from other aspects of the

debate, though these are far more tenuous. For example, the comparative

quickness and smoothness with which El responds to E,'s evidence may

suggest that El is already well familiar with E,'s "points" and has already

thought out counterarguments. If E,'s responsiveness to El ' s arguments

displays less quickness and smoothness, that may suggest that El ' s prior

mastery of the relevant information and support considerations exceeds that of

E,. Of course, quickness and smoothness are problematic indicators of infor-

mational mastery. Skilled debaters and well-coached witnesses can appear

better-informed because of their stylistic polish, which is not a true indicator

l 1 By "direct" justification I do not, of course, mean anything having to do with the basic-

ness of the conclusion in question, in the foundationalist sense of basicness. The distinc-

tion I am after is entirely different, as will shortly emerge. l 2 Edward Craig (1990: 135) similarly speaks of "indicator properties" as what an inquirer

seeks to identify in an informant as a guide to hisher truth-telling ability.


-- --

of superior expertise. This makes the proper use of indirect argumentative

justification a very delicate matter.13

To clarify the direcvindirect distinction being drawn here, consider two

different things a hearer might say to articulate these different bases of justifi-

cation. In the case of direct argumentative justifiedness, he might say: "In

light of this expert's argument, that is, in light of the truth of its premises

and the support they confer on the conclusion (both of which are epistemi-

cally accessible to me), I am now justified in believing the conclusion." In

indirect argumentative justifiedness, the hearer might say: "In light of the

way this expert has argued-her argumentative pe@omnce, as it were-I can

infer that she has more expertise than her opponent; so I am justified in infer-

ring that her conclusion is probably the correct one."

Here is another way to explain the direcvindirect distinction. Indirect

argumentative justification essentially involves an inference to the best

explanation, an inference that N might make from the performances of the

two speakers to their respective levels of expertise. From their performances,

N makes an inference as to which expert has superior expertise in the target

domain. Then he makes an inference from greater expertise to a higher proba-

bility of endorsing a true conclusion. Whereas indirect argumentative justifi-

cation essentially involves inference to the best explanation, direct argumen-

tative justification need involve no such inference. Of course, it might

involve such inference; but if so, the topic of the explanatory inference will

only concern the objects, systems, or states of affairs under dispute, not the

relative expertise of the contending experts. By contrast, in indirect argumen-

tative justifiedness, it is precisely the experts' relative expertise that consti-

tutes the target of the inference to the best explanation.

Hardwig (1985) makes much of the fact that in the novicelexpert situa-

tion, the novice lacks the expert's reasons for believing her conclusion. This

is correct. Usually, a novice (1) lacks all or some of the premises from which

an expert reasons to her conclusion, (2) is in an inferior position to assess the

support relation between the expert's premises and conclusions, and (3) is

ignorant of many or most of the defeaters (and "defeater-defeaters") that might

bear on an expert's arguments. However, although novice N may lack (all or

some of) an expert's reasons R for believing a conclusion p, N might have

reasons R* for believing that the expert has good reasons for believing p; and

N might have reasons R* for believing that one expert has better reasons for

believing her conclusion than her opponent has for hers. Indirect argumenta-

p~

" Scott Brewer (1998) discusses many of the same issues about novices and experts

canvassed here. He treats the present topic under the heading of novices' using experts'

"demeanor" to assess their expertise. Demeanor is an especially untrustworthy guide, he

points out, where there is a lucrative "market" for demeanor itself-where demeanor is

"traded" at high prices (1998: 1622). This practice was prominent in the days of the

sophists and is a robust business in adversarial legal systems.

96 ALVIN I. GOLDMAN

tive justification is one means by which N might acquire reasons R* without

sharing (all or any) of either experts' reasons R.I4 It is this possibility to

which Hardwig gives short shrift. I don't say that a novice in a novicef2-

expert situation invariably has such reasons R*; nor do I say that it is easy

for a novice to acquire such reasons. But it does seem to be possible.

4 . Agreement from Other Experts: The Question of Numbers

An additional possible strategy for the novice is to appeal to further experts.

This brings us to categories (B) and (C) on our list. Category (B) invites N to

consider whether other experts agree with El or with E,. What proportion of

these experts agree with El and what proportion with E,? In other words, to

the extent that it is feasible, N should consult the numbers, or degree of

consensus, among all relevant (putative) experts. Won't N be fully justified

in trusting El over E, if almost all other experts on the subject agree with E,,

or if even a preponderance of the other experts agree with El?

Another possible source of evidence, cited under category (C), also appeals

to other experts but in a slightly different vein. Under category (C), N should

seek evidence about the two rival experts' relative degrees of expertise by

consulting third parties' assessments of their expertise. If "meta-experts" give

El higher "ratings" or "scores" than E,, shouldn't N rely more on E, than E,?

Credentials can be viewed as a special case of this same process. Academic

degrees, professional accreditations, work experience, and so forth (all from

specific institutions with distinct reputations) reflect certifications by other

experts of El ' s and E,'s demonstrated training or competence. The relative

strengths or weights of these indicators might be utilized by N to distill

appropriate levels of trust for El and E, respecti~ely.'~

I treat ratings and credentials as signaling "agreement" by other experts

because I assume that established authorities certify trainees as competent

when they are satisfied that the latter demonstrate (1) a mastery of the same

methods that the certifiers deem fundamental to the field, and (2) knowledge

of (or belief in) propositions that certifiers deem to be fundamental facts or

laws of the discipline. In this fashion, ratings and conferred credentials

ultimately rest on basic agreement with the meta-experts and certifying

authorities.

When it comes to evaluating specific experts, there is precedent in the

American legal system for inquiring into the degree to which other experts

agree with those being evaluated.16 But precedented or not, just how good is

l4 Of course, in indirect argumentative justification the novice must at least hear some of

the expert's premises--or intermediate steps between "ultimate" premises and conclu-

sion. But the novice will not share the expert's jusrifiedness in believing those premises. I s These items fall under Kitcher's category of "unearned authority" (1993: 315). l 6 Appealing to other experts to validate or underwrite a putative expert's opinion--or,

more precisely, the basis for his opinion-has a precedent in the legal system's proce-


this appeal to consensus? If a putative expert's opinion is joined by the

consensual opinions of other putative experts, how much warrant does that

give a hearer for trusting the original opinion? How much evidential worth

does consensus or agreement deserve in the doxastic decision-making of a

hearer?

If one holds that a person's opinion deserves prima facie credence, despite

the absence of any evidence of their reliability on the subject, then numbers

would seem to be very weighty, at least in the absence of additional evidence.

Each new testifier or opinion-holder on one side of the issue should add

weight to that side. So a novice who is otherwise in the dark about the relia-

bility of the various opinion-holders would seem driven to agree with the

more numerous body of experts. Is that right?

Here are two examples that pose doubts for "using the numbers" to judge

the relative credibility of opposing positions. First is the case of a guru with

slavish followers. Whatever the guru believes is slavishly believed by his

followers. They fix their opinions wholly and exclusively on the basis of

their leader's views. Intellectually speaking, they are merely his clones. Or

consider a group of followers who are not led by a single leader but by a

small elite of opinion-makers. When the opinion-makers agree, the mass of

followers concur in their opinion. Shouldn't a novice consider this kind of

scenario as a possibility? Perhaps (putative) expert E, belongs to a doctrinal

community whose members devoutly and uncritically agree with the opinions

of some single leader or leadership cabal. Should the numerosity of the

community make their opinion more credible than that of a less numerous

group of experts? Another example, which also challenges the probity of

dures for deciding the admissibility of scientific expert testimony. Under the governing

test for admitting or excluding such testimony that was applicable from 1923 to 1993, the

scientific principle (or methodology) on which a proffered piece of testimony is based

must have "gained general acceptance in the particular field in which it belongs". (Frye

v. United States, 292 F. 1013 D.C. Cir. (1923)). In other words, appeal was made to the

scientific community's opinion to decide whether the basis of an expert's testimony is

sound enough to allow that testimony into court. This test has been superseded as the

uniquely appropriate test in a more recent decision of the Supreme Court (Daubert v.

Merrell Dow Pharmaceuticals, 509 U.S. 579 (1993)); but the latter decision also appeals

to the opinions of other experts. It recommends that judges use a combination of four

criteria (none of them necessary or sufficient) in deciding whether proffered scientific

expert testimony is admissible. One criterion is the old general acceptance criterion and

another is whether the proffered evidence has been subjected to peer review and publi-

cation. Peer review, obviously, also introduces the opinions of other experts. Of course,

the admissibility of a piece of expert testimony is not the same question as how heavily a hearer--e.g., a juror-should trust such testimony if he hears it. But the two are closely

intertwined, since courts make admissibility decisions on the assumption that jurors are

likely to be influenced by any expert testimony they hear. Courts do not wish to admit

scientific evidence unless it is quite trustworthy. Thus, the idea of ultimately going to the

opinions of other experts to assess the trustworthiness of a given expert's proffered testi-

mony is certainly a well-precedented procedure for trying to validate an expert's trust-

worthiness.

98 ALVIN I. GOLDMAN

greater numbers, is the example of rumors. Rumors are stories that are widely

circulated and accepted though few of the believers have access to the rumored

facts. If someone hears a rumor from one source, is that source's credibility

enhanced when the same rumor is repeated by a second, third, and fourth

source? Presumably not, especially if the hearer knows (or justifiably

believes) that these sources are all uncritical recipients of the same rumor.

It will be objected that additional rumor spreaders do not add credibility to

an initial rumor monger because the additional ones have no established relia-

bility. The hearer has no reason to think that any of their opinions is worthy

of trust. Furthermore, the rumor case doesn't seem to involve "expert"

opinions at all and thereby contrasts with the original case. In the original

case the hearer has at least some prior reason to think that each new speaker

who concurs with one of the original pair has some credibility (reliability).

Under that scenario, don't additional concurring experts increase the total

believability of the one with whom they agree?

It appears, then, that greater numbers should add further credibility, at least

when each added opinion-holder has positive initial credibility. This view is

certainly presupposed by some approaches to the subject. In the Lehrer-

Wagner (1981) model, for example, each new person to whom a subject

assigns "respect" or "weight" will provide an extra vector that should push

the subject in the direction of that individual's opinion.I7 Unfortunately, this

approach has a problem. If two or more opinion-holders are totally non-

independent of one another, and if the subject knows or is justified in believ-

ing this, then the subject's opinion should not be swayed--even a little-by

more than one of these opinion-holders. As in the case of a guru and his blind

followers, a follower's opinion does not provide any additional grounds for

accepting the guru's view (and a second follower does not provide additional

grounds for accepting a first follower's view) even if all followers are

precisely as reliable as the gum himself (or as one another)-which followers

must be, of course, if they believe exactly the same things as the guru (and

one another) on the topics in question. Let me demonstrate this through a

Bayesian analysis.

Under a simple Bayesian approach, an agent who receives new evidence

should update his degree of belief in a hypothesis H by conditioning on that

evidence. This means that he should use the ratio (or quotient) of two likeli-

hoods: the likelihood of the evidence occurring if H is true and the likelihood

l 7 Lehrer and Wagner say (p. 20) that one should assign somebody else a positive weight if

one does not regard his opinion as "worthless" on the topic in question-i.e., if one

regards him as better than a random device. So it looks as if every clone of a leader

should be given positive weight-arguably, the same weight as the leader himself, since

their beliefs always coincide-as long as the leader receives positive weight. In the

Lehrer-Wagner model, then, each clone will exert a positive force over one's own revi-

sions of opinion just as a leader's opinion will exert such force; and the more clones there

are, the more force in the direction of their collective opinion will be exerted.


of the evidence occurring if H is false. In the present case the evidence in

question is the belief in H on the part of one or more putative experts. More

precisely, we are interested in comparing (A) the result of conditioning on the

evidence of a single putative expert's belief with (B) the result of condition-

ing on the evidence of concurring beliefs by two putative experts. Call the

two putative experts X and Y, and let X(H) be X's believing H and Y(H) be

Y's believing H. What we wish to compare, then, is the magnitude of the

likelihood quotient expressed in (1) with the magnitude of the likelihood

quotient expressed in (2).

The principle we are interested in is the principle that the likelihood ratio

given in (2) is always larger than the likelihood ratio given in (I) , so that an

agent who learns that X and Y both believe H will always have grounds for a

larger upward revision of his degree of belief in H than if he learns only that

X believes H. At least this is so when X and Y are each somewhat credible

(reliable). More precisely, such comparative revisions are in order if the agent

is justified in believing these things in the different scenarios. I am going to

show that such comparative revisions are not always in order. Sometimes (2)

is not larger than (1); so the agent-if he knows or justifiably believes this-

is not justified in making a larger upward revision from the evidence of two

concurring believers than from one believer.

First let us note that according to the probability calculus, (2) is equiva-

lent to (3).

While looking at (3), return to the case of blind followers. If Y is a blind

follower of X, then anything believed by X (including H) will also be

believed by Y. And this will hold whether or not H is true. So,

and

100 ALVIN I. GOLDMAN

Substituting these two values into expression (3), (3) reduces to (1). Thus, in

the case of a blind follower, (2) (which is equivalent to (3)) is the same as

(I), and no larger revision is warranted in the two-concurring-believers case

than in the single-believer case.

Suppose that the second concurring believer, Y, is not a blind follower of

X. Suppose he would sometimes agree with X but not in all circumstances.

Under that scenario, does the addition of Y's concurring belief always provide

the agent (who possesses this information) with more grounds for believing

H? Again the answer is no. The appropriate question is whether Y is more

likely to believe H when X believes H and H is true than when X believes H

and H is false. If Y is just as likely to follow X's opinion whether H is true

or false, then Y's concurring belief adds nothing to the agent's evidential

grounds for H (driven by the likelihood quotient). Let us see why this is so.

If Y is just as likely to follow X's opinion when H is false as when it's

true, then (6) holds:

But if (6) holds, then (3) again reduces to (I), because the right-hand sides of

both numerator and denominator in (3) are equal and cancel each other out.

Since (3) reduces to (I), the agent still gets no extra evidential boost from

Y's agreement with X concerning H. Here it is not required that Y is certain

to follow X's opinion; the likelihood of his following X might only be 0.80,

or 0.40, or whatever. As long as Y is just as likely to follow X's opinion

when H is true as when it's false, we get the same result.

Let us describe this last case by saying that Y is a non-discriminating

reflector of X (with respect to H). When Y is a non-discriminating reflector of

X, Y's opinion has no extra evidential worth for the agent above and beyond

X's opinion. What is necessary for the novice to get an extra evidential boost

from Y's belief in H is that he (the novice) be justified in believing (6'):

If (6') is satisfied, then Y's belief is at least partly conditionally independent

of X's belief. Full conditional independence is a situation in which any

dependency between X and Y's beliefs is accounted for by the dependency of

each upon H. Although full conditional independence is not required to boost

N's evidence, partial conditional independence is req~ired . '~

We may now identify the trouble with the (unqualified) numbers principle.

The trouble is that a novice cannot automatically count on his putative

" I am indebted here to Richard Jeffrey (1992: 109-10). He points out that it is only condi-

tional independence that is relevant in these kinds of cases, not "simple independence"

defined by the condition: P(Y(H) / X(H) ) = P(Y(H)). If X and Y are even slightly reli-

able independent sources of information about H, they won't satisfy this latter condition.

EXPERTS: WHICH ONES SHOULD YOU TRUST?101

experts being (even partially) conditionally independent of one another. He

cannot automatically count on the truth of (6'). Y may be a non-discriminat-

ing reflector of X, or X may be a non-discriminating reflector of Y, or both

may be non-discriminating reflectors of some third party or parties. The same

point applies no matter how many additional putative experts share an initial

expert's opinion. If they are all non-discriminating reflectors of someone

whose opinion has already been taken into account, they add no further

weight to the novice's evidence.

What type of evidence can the novice have to justify his acceptance of (or

high level of credence in) (6')? N can have reason to believe that Y's route to

belief in H was such that even in possible cases where X fails to recognize

H's falsity (and hence believes it), Y would recognize its falsity. There are

two types of causal routes to Y's belief of the right sort. First, Y's route to

belief in H might entirely bypass X's route. This would be exemplified by

cases in which X and Y are causally independent eyewitnesses of the occur-

rence or non-occurrence of H; or by cases in which X and Y base their respec-

tive beliefs on independent experiments that bear on H. In the eyewitness

scenario X might falsely believe H through rnisperception of the actual event,

whereas Y might perceive the event correctly and avoid belief in H. A second

possible route to Y's belief in H might go partly through X but not involve

uncritical reflection of X's belief. For example, Y might listen to X's reasons

for believing H, consider a variety of possible defeaters of these reasons that

X never considered, but finally rebut the cogency of these defeaters and concur

in accepting H. In either of these scenarios Y's partly "autonomous" causal

route made him poised to avoid belief in H even though X believes it

(possibly falsely). If N has reason to think that Y used one of these more-or-

less autonomous causal routes to belief, rather than a causal route that guar-

antees agreement with X, then N has reason to accept (6'). In this fashion, N

would have good reason to rate Y's belief as increasing his evidence for H

even after taking account of X's belief.

Presumably, novices could well be in such an epistemic situation vis-8-

vis a group of concurring (putative) experts. Certainly in the case of concur-

ring scientists, where a novice might have reason to expect them to be criti-

cal of one another's viewpoints, a presumption of partial independence might

well be in order. If so, a novice might be warranted in giving greater eviden-

tial weight to larger numbers of concurring opinion-holders. According to

some theories of scientific opinion formation, however, this warrant could

not be sustained. Consider the view that scientists' beliefs are produced

entirely by negotiation with other scientists, and in no way reflect reality (or

Nature). This view is apparently held by some social constructionists about

science, e.g., Bruno Latour and Steve Woolgar (197911986); at least this is


- -

Kitcher's (1993: 165-66) interpretation of their view." Now if the social

constructionists are right, so interpreted, then nobody (at least nobody knowl-

edgeable of this fact) would be warranted in believing anything like (6').

There would never be reason to think that any scientist is more likely to

believe a scientific hypothesis H when it's true (and some other scientist

believes it) than when it's false (and the other scientist believes it). Since

causal routes to scientific belief never reflect "real" facts-they only reflect

the opinions, interests, and so forth of the community of scientists-(6') will

never be true. Anybody who accepts or inclines toward the indicated social-

constructionist thesis would never be justified in believing (6').20

Setting such extreme views aside, won't a novice normally have reason to

expect that different putative experts will have some causal independence or

autonomy from one another in their routes to belief? If so, then if a novice is

also justified in believing that each putative expert has some slight level of

reliability (greater than chance), then won't he be justified in using the

numbers of concurring experts to tilt toward one of two initial rivals as

opposed to the other? This conclusion might be right when all or almost all

supplementary experts agree with one of the two initial rivals. But this is

rarely the case. Vastly more common are scenarios in which the numbers are

more evenly balanced, though not exactly equal. What can a novice conclude

in those circumstances? Can he legitimately let the greater numbers decide the

issue?

This would be unwarranted, especially if we continue to apply the

Bayesian approach. The appropriate change in the novice's belief in H should

be based on two sets of concurring opinions (one in favor of H and one

against it), and it should depend on how reliable the members of each set are

and on how (conditionally) independent of one another they are. If the

members of the smaller group are more reliable and more (conditionally)

independent of one another than the members of the larger group, that might

imply that the evidential weight of the smaller group exceeds that of the

larger one. More precisely, it depends on what the novice is justified in

believing about these matters. Since the novice's justifiedness on these

matters may be very weak, there will be many situations in which he has no

distinct or robust justification for going by the relative numbers of like-

minded opinion-holders.

This conclusion seems perfectly in order. Here is an example that, by my

own lights, sits well with this conclusion. If scientific creationists are more

numerous than evolutionary scientists, that would not incline me to say that

a novice is warranted in putting more credence in the views of the former than

I y I myself interpret Latour and Woolgar as holding a more radical view, viz., that there is

no reality that could causally interact, even indirectly, with scientists' beliefs. 20 This is equally so under the more radical view that there are no truths at all (of a scien-

tific sort) about reality or Nature.


in the views of the latter (on the core issues on which they disagree). At least

I am not so inclined on the assumption that the novice has roughly compara-

ble information as most philosophers currently have about the methods of

belief formation by evolutionists and creationists respecti~ely.~' Certainly the

numbers do not necessarily outweigh considerations of individual reliability

and mutual conditional independence. The latter factors seem more probative,

in the present case, than the weight of sheer numbers.22

5 . Evidence fromInterests and Biases

I turn now to the fourth source of possible evidence on our original list:

evidence of distorting interests and biases that might lie behind a putative

expert's claims. If N has excellent evidence for such bias in one expert and no

evidence for such bias in her rival, and if N has no other basis for preferential

trust, then N is justified in placing greater trust in the unbiased expert. This

proposal comes directly from common sense and experience. If two people

give contradictory reports, and exactly one of them has a reason to lie, the

relative credibility of the latter is seriously compromised.

Lying, of course, is not the only way that interests and biases can reduce

an expert's trustworthiness. Interests and biases can exert more subtle distort-

ing influences on experts' opinions, so that their opinions are less likely to

be accurate even if sincere. Someone who is regularly hired as an expert

witness for the defense in certain types of civil suits has an economic interest

21 More specifically, I am assuming that believers in creation science have greater

(conditional) dependence on the opinion leaders of their general viewpoint than do

believers in evolutionary theory. 22 John Pollock (in a personal communication) suggests a way to bolster support for the use

of "the numbers". He says that if one can argue that P(X(H) / Y(H) & H) = P(X(H) / H),

then one can cumulate testimony on each side of an issue by counting experts. He further

suggests that, in the absence of countervailing evidence, we should believe that P(X(H) /

Y(H) & H) = P(X(H) 1 H). He proposes a general principle of probabilistic reasoning,

which he calls "the principle of nonclassical direct inference", to the effect that we are

defeasibly justified in regarding additional factors about which we know nothing to be

irrelevant to the probabilities. In Pollock (2000) (also see Pollock 1990) he formulates the

idea as follows. If factor C is irrelevant (presumably he means probabiiisticaiiy irrele-

vant) to the causal relation between properties B and A, then conjoining C to B should not

affect the probability of something's being A. Thus, if we have no reason to think that C

is relevant, we can assume defeasibly that P(Ax / Bx & Cx) = P(Ax / Bx). This principle

can.& applied, he suggests, to the case of a concurring (putative) expert. But, I ask, is it

generally reasonable for us--or for a novice-to assume that the opinion of one expert is

probabilistically irrelevant to another expert's holding the same view? I would argue in

the negative. Even if neither expert directly influences the opinion of the other, it is

extremely common for two people who work in the same intellectual domain to be influ-

enced, directly or indirectly, by some common third expert or group of experts. Interde-

pendence of this sort is widespread, and could be justifiably believed by novices. Thus,

probabilistic irrelevance of the sort Pollock postulates as the default case is highly ques-

tionable.


in delivering strong testimony in any current trial, because her reputation as a

defense witness depends on her present performance.

As a test of expert performance in situations of conflict of interest,

consider the results of a study published in the Journal of American Medical

Association (Friedberg et al., 1999). The study explored the relationship

between published research reports on new oncology drugs that had been

sponsored by pharmaceutical companies versus those that had been sponsored

by nonprofit organizations. It found a statistically significant relationship

between the funding source and the qualitative conclusions in the reports.

Unfavorable conclusions were reached by 38% of nonprofit-sponsored studies

but by only 5% of pharmaceutical company-sponsored studies.

From a practical point of view, information bearing on an expert's inter-

ests is often one of the more accessible pieces of relevant information that a

novice can glean about an expert. Of course, it often transpires that both

members of a pair of testifying experts have interests that compromise their

credibility. But when there is a non-negligible difference on this dimension, it

is certainly legitimate information for a novice to employ.

Pecuniary interests are familiar types of potential distorters of an individ-

ual's claims or opinions. Of greater significance, partly because of its greater

opacity to the novice, is a bias that might infect a whole discipline, sub-

discipline, or research group. If all or most members of a given field are

infected by the same bias, the novice will have a difficult time telling the real

worth of corroborating testimony from other experts and meta-experts. This

makes the numbers game, discussed in the previous section, even trickier for

the novice to negotiate.

One class of biases emphasized by feminist epistemologists involves the

exclusion or underrepresentation of certain viewpoints or standpoints within a

discipline or expert community. This might result in the failure of a commu-

nity to gather or appreciate the significance of certain types of relevant

evidence. A second type of community-wide bias arises from the economics

or politics of a sub-discipline, or research community. To advance its funding

prospects, practitioners might habitually exaggerate the probativeness of the

evidence that allegedly supports their findings, especially to outsiders. In

competition with neighboring sciences and research enterprises for both

resources and recognition, a given research community might apply compara-

tively lax standards in reporting its results. Novices will have a difficult time

detecting this, or weighing the merit of such an allegation by rival experts

outside the field.23

In a devastating critique of the mental health profession, Robyn Dawes (1994) shows that

the real expertise of such professionals is, scientifically, very much in doubt, despite the

high level of credentialism in that professional community.

EXPERTS: WHICH ONES SHOULD YOU TRUST? 10.5

23

6. Using Past Track Records

The final category in our list may provide the novice's best source of

evidence for making credibility choices. This is the use of putative experts'

past track records of cognitive success to assess the likelihoods of their

having correct answers to the current question. But how can a novice assess

past track records? There are several theoretical problems here, harking back

to matters discussed earlier.

First, doesn't using past track records amount to using the method of

(direct) "calibration" to assess a candidate expert's expertise? Using a past

track record means looking at the candidate's past success rate for previous

questions in the E-domain to which she offered answers. But in our earlier

discussion (section 2), I said that it's in the nature of a novice that he has no

opinions, or no confidence in his own opinions, about matters falling within

the E-domain. So how can the novice have any (usable) beliefs about past

answers in the E-domain by which to assess the candidate's expertise? In

other words, how can a novice, qua novice, have any opinions at all about

past track records of candidate experts?

A possible response to this problem is to revisit the distinction between

esoteric and exoteric statements. Perhaps not every statement in the E-domain

is esoteric. There may also be a body of exoteric statements in the E-domain,

and they are the statements for which a novice might assess a candidate's

expertise. But does this really make sense? If a statement is an exoteric

statement, i.e., one that is epistemically accessible to novices, then why

should it even be included in the E-domain? One would have thought that the

E-domain is precisely the domain of propositions accessible only to experts.

The solution to the problem begins by sharpening our esoteric/exoteric

distinction. It is natural to think that statements are categorically either

esoteric or exoteric, but that is a mistake. A given (timeless) statement is

esoteric or exoteric only relative to an epistemic standpoint or position. It

might be esoteric relative to one epistemic position but exoteric relative to a

different position. For example, consider the statement, "There will be an

eclipse of the sun on April 22, 2130, in Santa Fe, New Mexico." Relative to

the present epistemic standpoint, i.e., the standpoint of people living in the

year 2000, this is an esoteric statement. Ordinary people in the year 2000

will not be able to answer this question correctly, except by guessing. On the

other hand, on the very day in question, April 22, 2130, ordinary people on

the street in Santa Fe, New Mexico will easily be able to answer the question

correctly. In that different epistemic position, the question will be an exoteric

one, not an esoteric one.24 You won't need specialized training or knowledge

24 In the present discussion only epistemic esotericness, not semantlc esotericness, is in

question (see note 10).


to determine the answer to the question. In this way, the epistemic status of a

statement can change from one time to another.

There is a significant application of this simple fact to the expertlnovice

problem. A novice might easily be able to determine the truth-value of a

statement after it has become exoteric. He might be able to tell then that it is

indeed true. Moreover, he might learn that at an earlier time, when the state-

ment was esoteric for the likes of him, another individual managed to believe

it and say that it is (or would be) true. Furthermore, the same individual

might repeatedly display the capacity to assert statements that are esoteric at

the time of assertion but become exoteric later, and she might repeatedly turn

out to have been right, as determined under the subsequently exoteric circum-

stances. When this transpires, novices can infer that this unusual knower

must possess some special manner of knowing-some distinctive expertise-

that is not available to them. They presumably will not know exactly what

this distinctive manner of knowing involves, but presumably it involves

some proprietary fund of information and some methodology for deploying

that information. In this fashion, a novice can verify somebody else's exper-

tise in a certain domain by verifying their impressive track record within that

domain. And this can be done without the novice himself somehow being

transformed into an expert.

The astronomical example is just one of many, which are easily prolifer-

ated. If an automobile, an air-conditioning system, or an organic system is

suffering some malfunction or impairment, untrained people will often be

unable to specify any true proposition of the form, "If you apply treatment X

to system Y, the system will return to proper functioning." However, there

may be people who can repeatedly specify true propositions precisely of this

Moreover, that these propositions are true can be verified by novices,

because novices might be able to "watch" the treatment being applied to the

malfunctioning system and see that the system returns to proper functioning

(faster than untreated systems do). Although the truth of the proposition is an

exoteric matter once the treatment works, it was an esoteric matter before the

treatment was applied and produced its result. In such a case the expert has

knowledge, and can be determined to have had knowledge, at a time when it

was esoteric.26

It should be emphasized that many questions to which experts provide

answers, at times when they are esoteric, are not merely yeslno questions that

might be answered correctly by lucky guesses. Many of them are questions

that admit of innumerable possible answers, sometimes indefinitely many

25 They can not only recognize such propositions as true when others offer them; they can

also produce such propositions on their own when asked the question, "What can be done

to repair this system?" 26 I have discussed such cases in earlier writings: Goldman 1991 and Goldman 1999 (p.

269).


answers. Simplifying for purposes of illustration, we might say that when a

patient with an ailment sees a doctor, he is asking her the question, "Which

medicine, among the tens of thousands of available medicines, will cure or

alleviate this ailment?'Such a question is unlikely to be answered correctly

by mere guesswork. Similarly, when rocket scientists were first trying to

land a spaceship on the moon, there were indefinitely many possible answers

to the question, "Which series of steps will succeed in landing this (or some)

spaceship on the moon?" Choosing a correct answer from among the infinite

list of possible answers is unlikely to be a lucky guess. It is feats like this,

often involving technological applications, that rightly persuade novices that

the people who get the correct answers have a special fund of information and

a special methodology for deploying it that jointly yield a superior capacity

to get right answers. In this fashion, novices can indeed determine that others

are experts in a domain in which they themselves are not.

Of course, this provides no algorithm by which novices can resolve all

their two-expert problems. Only occasionally will a novice know, or be able

to determine, the track records of the putative experts that dispute an issue

before him. A juror in a civil trial has no opportunity to run out and obtain

track record information about rival expert witnesses who testify before him.

Nonetheless, the fact that novices can verify track records and use them to test

a candidate's claims to expertise, at least in principle and in some cases, goes

some distance toward dispelling utter skepticism for the novice12-expert situa-

tion. Moreover, the possibility of "directly" determining the expertise of a

few experts makes it possible to draw plausible inferences about a much

wider class of candidate experts. If certain individuals are shown, by the

methods presented above, to have substantial expertise, and if those individ-

uals train others, then it is a plausible inference that the trainees will them-

selves have comparable funds of information and methodologies, of the same

sort that yielded cognitive success for the original experts.27 Furthermore, to

the extent that the verified experts are then consulted as "meta-experts" about

the expertise of others (even if they didn't train or credential them), the latter

can again be inferred to have comparable expertise. Thus, some of the earlier

skepticism engendered by the novicel2-expert problem might be mitigated

once the foundation of expert verification provided in this section has been

established.

27 Of course, some experts may be better than others at transmitting their expertise. Some

may devote more effort to it, be more skilled at it, or exercise stricter standards in

credentialing their trainees. This is why good information about training programs 1s

certainly relevant to judgments of expertise.

108 ALVLN I. GOLDMAN

7. Conclusion

My story's ending is decidedly mixed, a cause for neither elation nor

gloom. Skeptical clouds loom over many a novice's epistemic horizons when

confronted with rival experts bearing competing messages. There are a few

silver linings, however. Establishing experts' track-records is not beyond the

pale of possibility, or even feasibility. This in turn can bolster the credibility

of a wider class of experts, thereby laying the foundation for a legitimate use

of numbers when trying to choose between experts. There is no denying,

however, that the epistemic situations facing novices are often daunting.

There are interesting theoretical questions in the analysis of such situations,

and they pose interesting practical challenges for "applied" social epistemol-

ogy. What kinds of education, for example, could substantially improve the

ability of novices to appraise expertise, and what kinds of communicational

intermediaries might help make the novice-expert relationship more one of

justified credence than blind trust.2a

References

Baron-Cohen, Simon (1995). Mindblindness. Cambridge, MA: MIT Press.

Bezuidenhout, Anne (1998). "Is Verbal Communication a Purely Preservative

Process?'Philosophical Review 107: 261-88.

Brewer, Scott (1998). "Scientific Expert Testimony and Intellectual Due

Process," Yale Law Journal 107: 1535-68 1.

Burge, Tyler (1993). "Content Preservation," Philosophical Review 102:

457-88.

Coady, C. A. J. (1992). Testimony. Oxford: Clarendon Press.

Craig, Edward (1990). Knowledge and the State of Nature-An Essay in

Conceptual Synthesis. Oxford: Clarendon Press.

Dawes, Robyn (1994). House of Cards: Psychology and Psychotherapy Built

on Myth. New York: Free Press.

Foley, Richard (1994). "Egoism in Epistemology," in F. Schmitt, ed.,

Socializing Epistemology. Lanham, MD: Rowman & Littlefield.

Friedberg, Mark et al. (1999). "Evaluation of Conflict of Interest in

Economic Analyses of New Drugs Used in Oncology," Journal of the

American Medical Association 282: 1453-57.

Gentzler, J. (1995). "How to Discriminate between Experts and Frauds: Some

Problems for Socratic Peirastic," History of Philosophy Quarterly 3:

227-46.

28 For helpful comments on earlier drafts, I am indebted to Holly Smith, Don Fallis, Peter

Graham, Patrick Rysiew, Alison Wylie, and numerous participants at the 2000 Rutgers

Epistemology Conference, the philosophy of social science roundtable in St. Louis, and

my 2000 NEH Summer Seminar on "Philosophical Foundations of Social Epistemology".


Goldman, Alvin (1991). "Epistemic Paternalism: Communication Control in

Law and Society," Journal of Philosophy 88: 11 3-3 1.

Goldman, Alvin (1999). Knowledge in a Social World. Oxford: Clarendon

Press.

Hardwig, John (1985). "Epistemic Dependence," Journal of Philosophy 82:

335-49.

Hardwig, John (1991). "The Role of Trust in Knowledge," Journal of Philos-

ophy 88: 693-708.

Jeffrey, Richard (1992). Probability and the Art of Judgment. New York:

Cambridge University Press.

Kitcher, Philip (1993). The Advancement of Science. New York: Oxford

University Press.

LaBarge, Scott (1997). "Socrates and the Recognition of Experts," in M.

McPherran, ed., Wisdom, Ignorance and Virtue: New Essays in Socratic

Studies. Edmonton: Academic Printing and Publishing.

Latour, Bruno and Woolgar, Steve (197911986). Laboratory Life: The

Construction of Scientific Facts. Princeton: Princeton University Press.

Lehrer, Keith and Wagner, Carl (1981). Rational Consensus in Science and

Society. Dordrecht: Reidel.

Pollock, John (1990). Nomic Probability and the Foundations of Induction.

New York: Oxford University Press.

Pollock, John (2000). "A Theory of Rational Action." Unpublished

manuscript, University of Arizona.

1 10 ALVIN I. GOLDMAN

Experts: Which Ones Should You Trust?kutrov.web.elte.hu/tudhat/11_goldman.pdf · 2014-05-15 ·...

Documents

Transcript of Experts: Which Ones Should You Trust?kutrov.web.elte.hu/tudhat/11_goldman.pdf · 2014-05-15 ·...